The paper evaluates the efficiency of multi-layer perceptron (MLP) learning algorithms, specifically focusing on back propagation methods for plant leaf classification. It compares incremental back propagation, Levenberg-Marquardt, and batch propagation techniques, concluding that batch training results in higher accuracy and faster processing. Various experiments demonstrated a linear increase in classification accuracy across different training methods, indicating the potential of MLP-based systems in identifying plant species through leaf image analysis.